Incorrect output for `expand()` when dealing with asymmetric trees

[abhsarma]

When dealing with asymmetric trees (see example below), the output of expand(M) is not accurate. For instance consider the following multiverse:

# a random simulated dataset
N = 100

df = tibble(
  score = rlnorm(N),
  pretest1 = round(runif(N) * 100),
  pretest2 = round(runif(N) * 100),
  reaction_time = rnorm(N, 5, 1),
  HR = round(rnorm(N, 90, 10)),
  fatigue = sample(seq(1:7), size = N, replace = TRUE)
)

fit = branch(model, 
         "normal" ~ glm(log(score) ~ pretest1 + pretest2 + 
                        branch(interaction, 
                               "yes" ~ reaction_time:HR,
                               "no" ~ reaction_time:HR) + 
                        fatigue, data = df, family = "gaussian"),
         "binomial" ~ glm(score ~ pretest1 + pretest2 + 
                          fatigue, data = df, family = "binomial")
      )

This multiverse consists of three distinct paths, but the table produced by expand() contains four. This is probably due to naively taking all possible combinations of options for each parameter

[abhsarma]

Not sure if there's a good solution for this. This is not a problem if assymetric trees are created through %when% or branch_assert, only when they are created through nested branch calls.
Currently, I am providing a warning message if there are asymmetric trees due to nested branch calls as in the example above.